Transfer Timing and the Length of Medication Weaning for Neonatal Opioid Withdrawal Syndrome

This investigation uses a single-site retrospective cohort design. Situated in an urban area of New Jersey, Children’s Specialized Hospital is a regional provider of coordinated multidisciplinary clinical services to children with special health care needs from birth through age 21 and collaborates widely with regional pediatric acute care providers. Before referral to this tertiary care facility with specialized NOWS care, infants are assessed at the birth hospital on the basis of the reported history of maternal substance use during pregnancy, as well as urine and meconium toxicology screening. The development of withdrawal symptoms is typically monitored in the birth hospital NICU or newborn nursery by using the modified Finnegan scoring system^6,21 . After severe NOWS is identified and pharmacotherapy treatment is initiated, the NOWS treatment team at the tertiary care facility receives the referral from the partner acute care facility to ensure a timely transition and continuity of care, with the goal of transferring the infant to tertiary care by DOL 14.

The infants receive the same weaning medication initiated in the acute care setting after transfer to tertiary care. The first-line pharmacotherapy is either morphine or methadone; clonidine or phenobarbital are used as second-line therapies in some cases. The monitoring of NOWS symptoms continues using Finnegan scores assessed at 4-hour intervals and pharmacotherapy is adjusted as follows. After admission, medication doses decrease by 10% of the initial dose for every 3 consecutive scores <8, or the average of 3 scores <24. The weaning schedule continues until the morphine dose is <0.04 mg/kg per day or the methadone dose is <0.05 mg/kg per day. This regimen is complemented by nonpharmacological support provided by a multidisciplinary clinical team from physical, occupational, and speech therapies, as well as nursing, child life, and psychology. The patients are treated in rooms with dimmed lights and low noise. When feasible, caregivers are involved in the infants’ care, which includes training on strategies for calming, feeding, sleeping, bonding, and well-baby care. In addition to medication weaning, the NOWS treatment goals include weight gain, growth, safe and efficient feeding, and fostering self-soothing abilities to promote neurotypical development. The benchmark for treatment duration is 2 to 4 weeks, on average, with all infants being medication-free at least 48 hours before discharge.

This study uses deidentified data abstracted from electronic medical records for all infants transferred for medication weaning for NOWS between May 2016 and June 2021. As a regional medical rehabilitation provider, Children’s Specialized Hospital received patients from 14 birth hospitals, and a total of 92 infants were admitted during the study period. Each infant had an International Classification of Diseases, 10th Revision code of P96.1 (i.e., neonatal withdrawal symptoms from maternal use of drugs of addiction) and was vetted for a primary diagnosis of NOWS by the clinically trained members of the study team. Abstractions were performed by 3 trained Research Assistants and reviewed by a clinical Research Coordinator for data quality. Five percent of the cases were reabstracted as an additional data quality check. Two cases were excluded because of the presence of a congenital medical condition that required the need for a longer NICU stay before transfer to the tertiary care facility. Three additional cases were excluded from the analysis: 1 infant with atypical age at admission on DOL 71, 1 infant treated with phenobarbital only, and 1 infant with a peak Finnegan score of 7 during the acute hospitalization phase, which was less than the typical score of 8 or higher reported in the literature as the symptom threshold for further NOWS monitoring.²¹  This yields a final sample of n = 87 infants with complete data for analysis. The data elements did not include personal health information, and therefore, the institutional review board determined that this investigation was exempt from review.

The primary outcome measure, length of weaning (LOW), is calculated as the number of days that elapsed from transfer to specialized NOWS care and the last day of pharmacotherapy, plus a 48-hour observation window. LOW is assessed for this analysis because additional time postmedication weaning was required for discharge disposition and caregiver training in many cases. The length of stay in both acute care (the number of days from birth to transfer) and tertiary care stay (the number of days between specialized hospital admission and discharge) is calculated and presented for comparison purposes.

The focal predictor variable is early (DOL 0–14) versus later (DOL 15 or more) transfer to tertiary care, created from the acute care length of stay. The recommended observation window for the development of withdrawal symptoms varies by type of opioid exposure, from 3 to 7 days,¹⁴  and the current protocol at the study site is to work with sending institutions to initiate the transfer of the infant for medication weaning between 7 and 14 days. The upper bound was selected to define the predictor variable for this analysis. Infant demographic characteristics include sex at birth and primary payer (Medicaid vs private). Infant race/ethnicity information was reported as underdetermined for more than one-third of the sample and, therefore, is excluded from the analysis. Infant growth and development are captured by measures of gestational age (weeks), birth weight (grams), a binary measure of small size for gestational age (1 = yes) adjusted for sex, maternal parity,²²  and head circumference (centimeters) assessed at admission to tertiary care. Several binary predictors captured birth information, including Cesarean (versus vaginal) delivery, the presence of any health complication in 1 or more bodily systems postbirth, such as noted respiratory, cardiac, neurologic, gastrointestinal, or sensory problems or birth injuries (1 = yes). NOWS severity is represented by the peak Finnegan score recorded after birth. Maternal characteristics include age at delivery in years and first birth (1 = yes).

In-utero narcotic exposures are classified as either maintenance therapy opioids only (i.e., either methadone or buprenorphine), opioid combination (i.e., primarily heroin combined with exposure to maintenance therapy opioids or prescription painkillers, such as Vicodin, codeine, or Oxycontin), or polysubstance (i.e., any opioid exposure with other nonnarcotics, such as cocaine, amphetamines, or nonprescription benzodiazepines or sedatives). Exposure to alcohol, tobacco, and cannabis is measured with separate, binary variables (1 = yes). Exposure to prescription psychotropic medications (1 = yes) is noted as a risk factor for NOWS, and therefore, a binary measure capturing maternal use of prescription psychotropic medication during pregnancy was created on the basis of the use of selective serotonin reuptake inhibitors, tricyclic, antianxiety, and antipsychotic medications. A measure of sexually transmitted infection (STI) exposure (1 = yes) is based on positive reports of prenatal exposure to 1 or more of the following conditions: Chlamydia, gonorrhea, hepatitis B, hepatitis C, herpes, HIV, syphilis, or HPV. A binary measure is used to assess the primary weaning medication used (1 = morphine, 0 = methadone). A small portion of the cases received second-line therapeutics (ie, clonidine, phenobarbital, or both). Because of the low-frequency counts of these cases, the primary weaning medication was used in the analysis.

χ² and independent sample t tests are used to test the differences between the early versus later admission groups in the distributions and means of the categorical and continuous variables, respectively. The outcome, LOW, is a positive integer with no 0 values. Therefore, 0 truncated negative binomial regression models are used to determine the relationship between the type of admission and LOW adjusted for relevant infant and maternal characteristics. An iterative, parsimonious modeling strategy first fits the model with admission type (early vs later) and then adjusts for key predictors associated with LOW. The model includes a priori predictors of NOWS from the literature, specifically, in-utero exposure to tobacco, psychotropic medications, and STI, as well as birth and maternal characteristics associated with the primary outcome. A P value of <.05 is used to assess the significance and a positive likelihood ratio test is used to compare differences in model fit with each additional covariate. The Akaike information criterion is used to assess overall model fit. Variables with low prevalence categories, such as alcohol exposure, are omitted from the models because small cell sizes limit model estimation and meaningful interpretation. All data analysis is conducted in Stata/SE version 17.²³ 

Table 1 summarizes the infant outcomes for the total sample and transfer timing. Of the 87 infants included in the analytic sample, 33 (38%) are early transfers (i.e., admitted to the specialized setting by DOL 14) and 54 (62%) are later transfers (i.e., admitted after DOL 15). LOW is an average of 13.6 days for the early group, whereas the latter group has an average LOW of 22.2 days (P < .001). The average tertiary care length of stay is 15.4 days versus 24.1 days (P < .001) for the early and later groups, respectively.

Table 1 also illustrates the distributions of infant, maternal, and exposure characteristics for the total sample and by transfer timing. The sample is 52% male, and the majority of the infants are Medicaid recipients (89%). The average maternal age is ∼30 years and, for 26%, this is a first birth. The majority of the infants in this sample are at full-term (i.e., born at >38 weeks gestation) and 25% are classified as small for gestational age. Prenatal care is noted for less than one-third of the sample and ∼40% were delivered by Cesarean section. Overall, health complications postbirth are common (72%), and the occurrence of complications did not differ between the transfer groups. The average peak Finnegan score noted after birth is ∼12, indicating that severe withdrawal symptoms are commonly observed shortly after birth among this sample. Overall, few differences are observed in the infant, maternal, birth, or exposure characteristics between the transfer groups. Of note, the early transfer group has a smaller average head circumference (mean: 34.1 vs mean: 35.2, P = .001), is less likely to have been exposed to an STI (44% vs 67%, P = .02), and more likely to be weaned using morphine (87.9% versus 68.5% P = .041).

Zero-truncated negative binomial regression models are used to estimate the difference in the LOW number of days by transfer timing adjusted for other salient risk factors (Table 2). The unadjusted incident rate ratio presented in Model 1 reveals that later transfer to a specialized NOWS treatment setting in this sample increases the LOW days by a factor of 63.7% (95% confidence interval [CI]: 1.30–2.07). The significant gap in LOW by transfer type persists, albeit at a reduced rate (35%, 95% CI: 1.08–1.68), after adjusting for infant sex, birth weight, small for gestational age, postbirth health complications, maximum Finnegan score, STI exposure, maternal age, and exposures to different opioid types, psychotropic medications, and tobacco in-utero (Model 2). The estimates presented in the fully adjusted model (Model 2) also indicate that the LOW days increase by a factor of 5% (95% CI: 1.01–1.08), with every 1-point increase in the maximum Finnegan score recorded postbirth, and exposure to psychotropic medications in-utero increases LOW by a factor of 39% (95% CI: 1.11–1.75), holding all other covariates constant. There are no significant differences estimated in LOW days by type of opioid exposure (ie, maintenance therapy only, opioid combination, or polysubstance), or the other infant, maternal, and exposure covariates.

The final model is restricted to include the predictors observed to increase the incident rate in the LOW days and opioid type (Model 3). This parsimonious model estimates that later transfer to NOWS treatment increases the LOW days by a factor of 47% (95% CI: 1.18–1.82), controlling for differences in NOWS severity, type of opioid exposure, and exposure to psychotropic medication. The Akaike information criterion in Table 2 suggests that the restricted model fits the data best (Model 3). Figure 1 illustrates both the unadjusted and adjusted difference from Model 3 in the predicted counts of LOW days between the early and later transfer. The adjusted predicted number of LOW days is 14.2 among infants in the early transfer group, or ∼6.6 days less than the later transfer group (20.8 days).

This investigation reveals that early referral to an inpatient tertiary care setting that provides integrated pharmacological and nonpharmacological treatment with environmental support for NOWS substantially decreases the LOW. Specifically, infants who transfer from the NICU to tertiary care by DOL 15 or later require almost an additional week of treatment compared with infants who transfer on or before DOL 14. This gap persists after controlling for NOWS severity and additional exposures. These findings are consistent with previous studies revealing improved clinical outcomes among infants receiving multidisciplinary NOWS care on low-stimulation inpatient units^13,15,19,20  Because many newborns with severe NOWS receive pharmacotherapy for withdrawal that may eventually require referral to an inpatient unit for medication weaning supplemented with environmental supports, the earlier timing of referrals may reduce the length of stay and minimize the other adverse effects of NOWS treatment on infants and families, such as prolonged medication exposure, separation from caregivers, and additional health care costs.

The peak Finnegan score recorded postbirth and prenatal exposure to psychotropic medication also increase LOW. This is consistent with previous reports indicating that NOWS symptoms are influenced by multiple factors. In particular, psychotropic medication increases the risk of severe withdrawal symptoms in infants.^1,10,11  When observed in combination with opioid exposure, these indicators may provide important markers to help identify infants at high risk for developing severe NOWS and in need of early referral to a multidisciplinary inpatient treatment program.

This study focuses on LOW, which is a departure from previous reports focused on overall LOS. Although an important clinical outcome, overall, LOS offers less precision because this outcome may encompass additional hospital days that accrue from the discharge disposition challenges rather than the need for further NOWS treatment. Nonetheless, the infants in this sample have notably lengthier hospitalizations compared with previous reports revealing that the average LOS for NOWS treatment ranges from 6 to 23 days.^3,19,24  The infants in our sample have an average LOS of 18.9 days in acute care and 20.8 days in tertiary care, which is consistent with other evidence suggesting that the duration of hospitalization and treatment (31.5 days and 25.3 days) is significantly longer for infants identified with high-severity versus low-severity NOWS.²⁵  This investigation does capture multiple rather than single hospitalizations and is methodologically distinct from previous reports. Notably, the authors of previous reports largely estimate overall LOS using aggregate insurance claims or inpatient databases, which may capture a wider range of opioid exposures from time-limited, licit opioid use to sustained, illicit use of prescription drugs and other narcotics. Although this study was not able to confirm a diagnosis of current or past maternal substance use disorder, the opioid exposure characteristics of the sample suggest that this study captures NOWS resulting from more extensive maternal drug use and additional risk factors, such as the increased risk for psychiatric comorbidities in this population.

This study adds to the growing evidence base by revealing that the early timing of transfer to a pediatric inpatient unit that is equipped to provide nonpharmacological and environmental support is critical and results in better outcomes for these patients. However, the authors of this investigation could not assess the factors related to the initiation of pharmacotherapy in the NICU settings and how variation in the initial clinical management of NOWS potentially contributes to the timing of transfer to a pediatric inpatient setting for medically supervised weaning. Although it is recommended that nonpharmacological treatment initially be provided to a newborn exposed to opioids, several sources note that pharmacotherapy may be necessary when withdrawal symptoms are severe and unmitigated by nonpharmacological therapies,⁷  and evidence suggests that pharmacotherapy protocols are widely used.^26,27  For instance, Snowden et al found that 96% of NICUs and 64% of non-NICUs in participating medical centers surveyed in 2017 in the United States reported having a standard protocol for pharmacological treatment of NOWS, and 73% of NICUs and 71% of non-NICUs used the modified Finnegan scoring to aid in NOWS assessment and management.¹  A 2020 multisite study conducted by Young and colleagues revealed that of 1377 infants with NOWs born between July 2016 to June 2017, 48.3% received pharmacological treatment, and all infants were assessed for withdrawal severity by the Finnegan scoring system. These reports provide just a few recent examples of the considerable variation that persists in the identification and initial management of NOWS, despite recommendations to integrate nonpharmacological approaches and functional assessment in NOWS treatment protocols.¹⁴  Moreover, there are ongoing issues, such as the need for updated, standardized assessment tools, written protocols, and provider education that would be important to assess to understand variation in pharmacological treatment initiation, transfer, and subsequent medication weaning as part of a future investigation.

By investigating the effect of timing, this study expands the knowledge base to support practitioners’ decision-making on transfer timing, which was currently determined by an array of medical, administrative, and social factors. Although the NOWS symptoms typically develop between 3 and 7 days of birth other sources note that symptoms may develop weeks after birth and last for a few months.⁵  The timeline for transfer for medication weaning and further treatment, particularly for severe NOWS, is less well-known and in need of further investigation. There are substantial practice variations in the initial risk assessment and observation period for NOWS at referring hospitals, which can average several days for infants exposed to long-acting opioids.²⁸  In addition, the provision of nonpharmacological supports and policies for parental involvement, such as rooming-in, caregiving, or breastfeeding differ across these various facilities. A detailed description of each health care setting is beyond the scope of this report, but it would be important to understand institutional variation as part of a future investigation because these factors are also likely to contribute to transfer timing and length of weaning. Administrative factors related to interinstitutional referrals and social factors related to the custodial situation of the child may also increase the time involved to complete the transfer of these infants from a NICU or nursery setting to tertiary care. Ideally, NOWS infants would be identified and transferred to care settings that are equipped to provide the combination of treatment and environmental supports that are unique to this patient population even earlier in the postnatal period, thereby fostering positive clinical outcomes (i.e., decreased length of weaning, length of stay, and caregiver separation) and early infant development. Finally, the need to improve education and communication between providers, parents/guardians, and, in many cases, child protective services, about the benefit of treatment on a pediatric inpatient unit is a consistent barrier identified at our facility that could also be targeted to expedite transfer and medication weaning.

This investigation has several acknowledged limitations. The interpretability of the results is qualified by a small, nonrepresentative sample. The use of data collected from a single site also limits the generalizability of these findings and comparability to patterns reported in population-based studies. The manual abstraction of free-text data elements from electronic medical charts can be subject to reliability concerns. These concerns are mitigated by using an interrater approach with oversight provided by clinical specialists with extensive knowledge of the patient population. There are data elements that would have informed the modeling strategy and enhanced comparability to the previous literature but were not consistently available, such as the duration of medication exposure at the referring (i.e., birth) hospital, the maximum dosage received before transfer, infant and maternal race and ethnicity, and maternal education level. Information on prenatal exposures is determined from a combination of toxicity screening data and maternal self-report. The challenging life circumstances and stigma that accompanies parental addiction likely introduce reporting bias on specific prenatal exposures. Finally, data specific to maternal drug use history such as the timing during pregnancy, duration, and dosages were not available.

To our knowledge, few studies have investigated the relationship between clinical outcomes and the timing of transfer from acute to tertiary care settings for multidisciplinary NOWS care with environmental support. NOWS is an evolving health threat to children, and the scale of this problem presents a challenge for developing consensus on standard practice techniques for assessing NOWS risk and implementing care that includes nonpharmacological therapy and environmental support. Given the wide variability in NOWS management across health care settings, this study provides evidence that can be used to inform practice guidelines for the timing of referral for infants with severe and persistent NOWS symptoms. Moreover, this study adds to the growing evidence that concurrent exposure to opioids and psychotropic medication may increase the risk for severe NOWS symptoms, providing an indication for a high-risk subgroup in need of evidence-based early intervention that is both pharmacologic and nonpharmacologically based. In addition, collaborative, multisite studies with representative samples are needed to produce generalizable evidence that can be used to understand differences between NOWS infants who are discharged from the hospital and those who are referred to tertiary care. The results of further investigation could then be translated to improve care coordination and educational efforts across health care settings, disciplines, and providers. Further research that assesses the complex administrative, family, and social reasons that affect transfer timing would also benefit care coordination for this fragile and high-risk population. The findings from this research may be used to improve communication between health care institutions and inform clinical guidelines aimed at decreasing medication use, hospital stays, and separation from caregivers for this vulnerable population.

The members of the study team not included in authorship are as follows: Ms Kristen Naples, MPH, Ms Stephanie Jimenez, MSW, Ms Alexa Bartalotta, BS, Ms Ann Kutlik, BA, and Ms Maayan Malomet, BA.

We would like to thank Ms Naples, who initially helped the research team acquire the cohort data, Ms Jimenez, who coordinated data collection and conducted data quality control checks, and Ms Bartalotta, Ms Kutlik, and Ms Malomet for their contributions to data collection.